A randomized trial of risperidone, placebo, and haloperidol for behavioral symptoms of dementia

Protocol.

Patients were eligible for inclusion in the trial if they were at least 55 years of age, were institutionalized, and had a diagnosis of primary degenerative dementia of the Alzheimer type, vascular dementia, or mixed dementia according to the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV).3 Eligible patients had scores of ≥4 on the Functional Assessment Staging (FAST); ≤23 on the Mini-Mental State Examination (MMSE); >1 on the Behavior Pathology in Alzheimer’s Disease Rating Scale (BEHAVE-AD) global rating; and ≥8 on the BEHAVE-AD total score.25-27

Exclusion criteria included other conditions that diminish cognitive function; other psychiatric disorders; clinically relevant organic or neurologic disease; ECG or laboratory abnormalities; administration of a depot neuroleptic within one treatment cycle of Visit 1; history of allergic reaction to neuroleptics or history of neuroleptic malignant syndrome; or participation in clinical trial(s) with investigational drugs during the 4 weeks preceding this trial.

The protocol was reviewed by an independent ethics committee or local institutional review board. The study was performed in accordance with the Declaration of Helsinki and its subsequent revisions. Subjects or their guardians gave written informed consent before participation.

The double-blind phase was preceded by a 1-week single-blind washout phase during which all psychotropic medications were discontinued (minimum duration, 3 days). The 12-week active treatment period (risperidone or haloperidol) started with 0.25 mg (1 mg/mL oral solution), with increments of 0.25 mg every 4 days, if indicated, to 1 mg twice daily. If the patient reached 1 mg twice daily without a sufficient therapeutic response and no signs of EPS, the dose could be increased to a maximum of 2 mg twice daily. The dose could be decreased at any time if significant adverse events occurred or if the investigator judged a decrease was warranted. Concomitant use of antipsychotics, antidepressants, lithium, carbamazepine, and valproic acid was not permitted during the study. Lorazepam was permitted if limited to 4 days per week for the first 4 weeks of double-blind treatment. If it was needed beyond week 4, the patient was to discontinue study participation. Patients could be withdrawn from the trial at the investigator’s discretion if a serious adverse event occurred, if the randomization code was broken, or if consent was withdrawn.

Patients were evaluated at selection, baseline, and weeks 1, 2, 4, 6, 8, 10, and 12. Efficacy assessments included the BEHAVE-AD, Cohen-Mansfield Agitation Inventory (CMAI), and Clinical Global Impression (CGI).28 The BEHAVE-AD is a 25-item scale that measures behavioral symptoms in seven clusters (paranoid and delusional ideation; hallucinations; activity symptoms; aggressiveness; diurnal rhythm symptoms; affective symptoms; and anxieties and phobias) scored on a four-point scale of increasing severity. The aggressiveness subscale score is the sum of three symptom scores (table 1). The developers of the scale made minor adaptations for use in an institutional setting. Four items were modified under the Paranoid and Delusion Ideation (item 2—one’s house is not one’s home delusion; 3—caregiver (or nurse or nursing aide) is an impostor delusion; 4—delusion of abandonment; and 5—delusion of infidelity). The BEHAVE-AD global rating score was based on a four-point scale of increasing severity as follows: 0, behavior not disturbing or dangerous to patient or environment; 1, behavior mildly disturbing or dangerous to patient or environment; 2, behavior moderately disturbing or dangerous to patient or environment; and 3, behavior severely disturbing or dangerous to patient or environment.

The CMAI is an established rating scale used in nursing homes to assess 29 agitated behaviors on a seven-point scale of increasing frequency. CMAI cluster scores include physical, verbal, and total aggression and physical, verbal, and total nonaggressive scores.29 CGI rating by the investigator measured behavioral symptoms on a seven-point scale of increasing severity. To minimize variables such as location and caregiver influences on outcome, all patients were institutionalized and were to stay on the same ward for the duration of the trial. Information was obtained from the nursing staff by the investigators who were trained in the use of the BEHAVE-AD and CMAI through videotapes prepared by authors of the scales.

The primary endpoint was defined as the percentage of patients with a 30% or greater improvement (from baseline to endpoint) on the BEHAVE-AD total score. To detect a 20% difference in the response rate at the 5% two-sided significance level with 80% power, a total of 91 patients was required in each treatment group (risperidone and placebo). Therefore, 114 patients were to be enrolled to each group to allow for a 20% dropout rate. Secondary parameters included the shift versus baseline in the BEHAVE-AD total score and global rating, the CMAI total score, the MMSE total score, and the FAST, as well as the CGI and the number of early withdrawals owing to inefficacy. Because aggression was the most dominant symptom, an additional analysis of the shift versus baseline was performed on categorical responses for aggression cluster scores.

Tolerability assessments included the Extrapyramidal Symptom Rating Scale (ESRS), a 55-item rating scale to assess the severity of EPS.30 Cognitive function was measured by the MMSE, an 11-item rating scale with scores ranging from 0 to 30; and functional status was assessed by the FAST, a seven-stage scale of increasing levels of disability. Sedation was rated on a seven-point scale of increasing severity. Additional assessments included reports of adverse events and concomitant medication use, as well as vital signs and electrocardiogram.

Analyses were performed according to the intent-to-treat principle (all randomized patients who received at least one dose of study medication and who had at least one post-baseline assessment). Between-group comparisons were two-tailed and determined at both endpoint (the last evaluation for each patient) and week 12 (observed case analysis, completers). The risperidone and placebo groups were compared on measures of efficacy and tolerability. As defined by the protocol, statistical comparisons between the haloperidol, risperidone, and placebo groups were made for tolerability analyses but not efficacy analyses. A post hoc analysis compared haloperidol with risperidone and placebo for efficacy outcomes.

For continuous measures (BEHAVE-AD total and cluster scores, CMAI total and cluster scores, MMSE, and FAST), the change from baseline score was analyzed using analysis of covariance (ANCOVA) models with factors for treatment, country, the treatment country interaction, and baseline score as covariates. If the treatment country interaction was not significant at the 10% level, the interaction term was removed from the model. The Least Squares (LS) means for treatment were calculated based on the reduced model.

BEHAVE-AD total scores were compared using the Cochran-Mantel-Haenszel (CMH) test for general association, controlling for country. CGI ratings were compared using the CMH row mean score test, controlling for country. The numbers of patients who discontinued prematurely because of lack of efficacy were compared using the CMH test for general association, controlling for country.

EPS severity was measured by the shift from baseline to endpoint in the ESRS total score (parkinsonism + dystonia + dyskinesia). Between-group comparisons were made using the same ANCOVA model as described earlier. Sedation data were analyzed using the Van Elteren test, controlling for country. The incidence of adverse events and the number of dropouts because of adverse events were compared using the CMH test for general association, controlling for country. Examination results, vital signs, weight, and electrocardiogram and laboratory data were analyzed descriptively. Frequencies of clinically important deviations from baseline and normal ranges were reported.

Assignment.

After initial screening and a 1-week, single-blind, placebo washout period, eligible patients were randomly assigned to treatment with risperidone, haloperidol, or placebo according to a predefined randomization code generated by the Janssen Research Foundation, Beerse, Belgium. Balancing ensured that an equal number of subjects were allocated to each treatment group at each center. Patient numbers were to be assigned in consecutive order.

Masking.

Trial medication was blinded and randomized using a computer-generated code and packaged as an oral solution containing placebo, 1 mg/mL risperidone, or 1 mg/mL haloperidol. All solutions were labeled with the trial number, patient number, and expiration date. One blister pack containing 20 placebo tablets was provided for each patient for the single-blind washout phase. For the double-blind phase, 7 × 100 mL bottles were prepared for each patient containing one of the three treatments (one bottle used per visit).

Participant flow and follow-up.

The trial was conducted from March 1995 to December 1996 at 51 centers in 8 countries. A total of 371 patients entered the 1-week washout period (figure 1). Twenty-seven patients dropped out during the washout period and never received double-blind treatment; they were not included in demographic, efficacy, or tolerability analyses. Three hundred forty-four patients completed washout and were randomly assigned to double-blind treatment with risperidone (n = 115), haloperidol (n = 115), or placebo (n = 114); they were included in all analyses. A total of 194 (56%) patients were women; median age was 81 years (range 56 to 97 years); the median time from the onset of dementia to trial entry was 4.3 years (range 79 days to 22.6 years); and the median duration of institutionalization was 4 months. A total of 229 (67%) patients had Alzheimer’s dementia according to the DSM-IV, 90 (26%) had vascular dementia, and 25 (7%) had mixed dementia. A total of 286 (85.9%) had received neuroleptics or other psychotropic drugs. The three study groups were comparable at baseline with respect to demographic and disease characteristics (table 2). The number of patients previously treated with high doses of antipsychotics (equivalent to >2 mg/d of haloperidol) was 12, 13, and 9 patients in the risperidone, haloperidol, and placebo groups.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Trial profile showing the flow of patients from recruitment through study completion.

Baseline assessments for efficacy and tolerability measurements were also comparable across the three groups (see table 2). Overall, these patients had severe dementia (FAST stage 6 in 210 [61%]; stage 7 in 106 [30.8%]) and poor cognitive function (mean MMSE scores of 7.9 to 8.8). The overall mean BEHAVE-AD total score at baseline was 16.5. The relative contribution of the BEHAVE-AD cluster scores at baseline shown in figure 2 indicates that aggressiveness was the dominant symptom in these patients. The mean aggressiveness cluster score was 4.9. The aggressiveness cluster contributes 30% of the symptoms to the BEHAVE-AD total score, yet it constitutes only 12% of the BEHAVE-AD scale. The “paranoid and delusional ideation” and “hallucination” clusters had the lowest proportional baseline scores (17% and 6% of the possible total, respectively) (see figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Relative contribution of Behavior Pathology in Alzheimer’s Disease Rating Scale (BEHAVE-AD) cluster or item scores at baseline (total cluster score divided by maximum cluster score possible): percentage of the maximal score for each cluster or item.

Sixty-six protocol deviations were recorded in 65 patients (use of disallowed concomitant medications in 64; exceeding maximum allowed dose of the trial medication in 2).

A total of 223 (68 risperidone, 81 haloperidol, and 74 placebo) completed the trial. The most common reasons for premature discontinuation were adverse events (n = 61, 50.4% of discontinuations) and lack of efficacy (n = 53, 43.8% of discontinuations); there were no significant between-group differences. Mean BEHAVE-AD total baseline scores were compared for dropouts and completers. Completers’ scores were 16.9, 16.7, and 16.6 in the risperidone, haloperidol, and placebo groups, respectively. Early dropouts (weeks 1 to 2) receiving risperidone had higher baseline scores than those receiving haloperidol or placebo (19.7, 16.8, and 15.0, respectively). The mean duration of double-blind treatment was 62.5 days, 70.3 days, and 67.6 days for patients in the risperidone, haloperidol, and placebo groups, respectively.

The mean dose at endpoint was 1.1 mg/d in the risperidone group and 1.2 mg/d in the haloperidol group. At endpoint, 111 (96.5%) patients in the risperidone group received a dose ≤2 mg/d and 4 (3.5%) received ≥2 mg/d.

Analyses.

The percentage of patients with clinical improvement (at least 30% reduction from baseline to endpoint in BEHAVE-AD total score) in the risperidone, haloperidol, and placebo groups was 54%, 63%, and 47%, respectively, at endpoint; and 72%, 69%, and 61%, respectively, at week 12 (p = 0.25 and p = 0.13, respectively, risperidone versus placebo).

Risperidone was associated with a greater reduction in both the severity and frequency of behavioral symptoms than was placebo (score reductions indicate less severe or fewer symptoms) (table 3, figures 3 and 4⇓). Differences between the risperidone and placebo groups were tested for statistical significance at endpoint and week 12. The risperidone group showed significantly greater improvement in the mean BEHAVE-AD total score at week 12 than did the placebo group (see table 3, figure 3). The therapeutic effect was most evident for aggressive symptoms. Risperidone was associated with a significantly greater improvement than was placebo in the BEHAVE-AD aggression cluster and CMAI aggressive scores (total, physical, and verbal cluster) at both endpoint and week 12 (see table 3, figures 3 and 4⇓). The superior effect of risperidone was seen as early as week 2 of treatment and was maintained over the next 10 weeks. Baseline scores for the “paranoid and delusional ideation” and “hallucination” clusters were quite low, and there were no significant changes with treatment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Mean BEHAVE-AD score shifts from baseline at each time point: (A) total, (B) aggressive. Risperidone versus placebo comparisons were two-tailed and determined at week 12 and at endpoint. *Significant change from baseline (see table 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Mean Cohen-Mansfield Agitation Inventory (CMAI) total aggressiveness score shifts from baseline at each time point. Risperidone versus placebo comparisons were two-tailed and determined at week 12 and at endpoint. *Significant change from baseline (see table 3).

Patients receiving risperidone also showed greater global improvement (CGI severity rating) than the placebo group at both endpoint (mean shifts: −0.7 and −0.4, respectively) and week 12 (−1.2 and −0.9, respectively) ( p < 0.05, CMH row mean score test for both comparisons). BEHAVE-AD global rating scores for study completers are shown in table 4. Both active treatment groups had more patients with no symptoms or mild symptoms compared with the placebo group. The placebo and haloperidol groups had more patients with severe symptoms than the risperidone group.

A subanalysis was performed excluding patients with vascular dementia. The results (mean shifts on BEHAVE-AD and CMAI scores) were consistent with those reported for all patients. At endpoint and week 12, there were greater improvements with risperidone than placebo on BEHAVE-AD aggressiveness score (p = 0.03 and p = 0.00, respectively) and CMAI total aggressiveness scores (p = 0.04 and p = 0.01, respectively).

A post hoc analysis compared haloperidol to risperidone and placebo on BEHAVE-AD and CMAI change scores (see table 3). Greater improvements were noted for haloperidol than placebo on BEHAVE-AD total and aggression cluster scores (p = 0.01) at endpoint. Also, there were significantly more improvements with risperidone than haloperidol on the BEHAVE-AD aggressiveness score (p = 0.05) and the CMAI total and verbal aggressive scores (p = 0.02 and p = 0.01, respectively) at week 12.

To determine whether somnolence or sedation influenced the outcome of treatment on behavior symptoms, a post hoc analysis excluded all patients who reported somnolence at any time during double-blind treatment. Results of this analysis (n = 302 patients; 100 risperidone, 109 placebo, 93 haloperidol) showed that risperidone was associated with a significantly greater reduction in behavioral symptoms than was placebo, indicating that the effect of risperidone is not attributable to sedation.

EPS severity was measured by the LS mean shifts from baseline in the ESRS total scores (parkinsonism + dystonia + dyskinesia). The severity of EPS at endpoint was not significantly different in the risperidone and placebo groups (−0.3 and −1.4, respectively); it was significantly greater with haloperidol than with risperidone (+1.6 and −0.3, respectively; p < 0.05). (Increased score indicates more severe symptoms.) The incidence of EPS-like adverse events was not significantly different in patients receiving risperidone (15%) or placebo (11%); it was significantly higher in those receiving haloperidol (22%) than placebo (p = 0.023).

Adverse events were reported by 88 (76.5%), 83 (72.8%), and 92 (80%) patients in the risperidone, placebo, and haloperidol groups, respectively. Adverse events occurring in ≥10% of patients in any one group were fall, injury, agitation, somnolence, and purpura (bruises caused by injuries or falls). Of these, only somnolence occurred in more patients receiving active treatment than in patients receiving placebo (haloperidol, 19 [18.3%]; risperidone, 14 [12.2%]; and placebo, 5 [4.4%]). There were no significant between-group differences in the occurrence of serious or severe adverse events.

Lorazepam use was similar among the three groups (risperidone, n = 34, 30%; haloperidol, n = 33, 29%; and placebo, n = 31, 27%). Overall sedation ratings, as measured by a seven-point scale of increasing severity, remained low throughout the trial. There was a slight increase from baseline in the risperidone (0.5 versus -0.1 placebo) and haloperidol (0.5) groups that was not considered to be clinically relevant. Endpoint scores rated sedation in all three groups between “not present” and “very low” (risperidone 1.8; placebo 1.4; haloperidol 1.9). MMSE scores at endpoint showed no cognitive deterioration with risperidone. LS mean changes were not significantly different between the risperidone and placebo groups but were significant (p < 0.05) between the haloperidol and placebo groups (risperidone −0.5; placebo 0.5; haloperidol −2.1), suggesting cognitive deterioration with haloperidol. Additionally, there was no significant difference between the three treatments on FAST ratings, indicating no negative effects on daily functioning. There were no consistent changes or clinically relevant abnormalities in vital signs (blood pressure and heart rate), laboratory safety parameters, body weight, or electrocardiogram.